Transport in Porous Media (2020) 135:39–58

https://doi.org/10.1007/s11242-020-01468-z

New Semi‑Analytical Solutions for Advection–Dispersion

Equations in Multilayer Porous Media

Elliot J. Carr1 

Received: 1 July 2020 / Accepted: 12 August 2020 / Published online: 1 September 2020
© Springer Nature B.V. 2020

Abstract
A new semi-analytical solution to the advection–dispersion–reaction equation for model-
ling solute transport in layered porous media is derived using the Laplace transform. Our
solution approach involves introducing unknown functions representing the dispersive flux
at the interfaces between adjacent layers, allowing the multilayer problem to be solved sep-
arately on each layer in the Laplace domain before being numerically inverted back to the
time domain. The derived solution is applicable to the most general form of linear advec-
tion–dispersion–reaction equation, a finite medium comprising an arbitrary number of lay-
ers, continuity of concentration and dispersive flux at the interfaces between adjacent lay-
ers and transient boundary conditions of arbitrary type at the inlet and outlet. The derived
semi-analytical solution extends and addresses deficiencies of existing analytical solutions
in a layered medium, which consider analogous processes such as diffusion or reaction–
diffusion only and/or require the solution of complicated nonlinear transcendental equa-
tions to evaluate the solution expressions. Code implementing our semi-analytical solution
is supplied and applied to a selection of test cases, with the reported results in excellent
agreement with a standard numerical solution and other analytical results available in the
literature.

Keywords  Advection dispersion reaction · Analytical solution · Layered media · Laplace

transform

1 Introduction

Solute transport in porous media due to dispersion and groundwater flow is typically
modelled using the advection–dispersion–reaction equation (van Genuchten and Alves
1982; Leij et al. 1991; Liu et al. 1998; Goltz and Huang 2017). While solving this equa-
tion in heterogeneous porous media usually requires the application of numerical meth-
ods, analytical solutions are generally preferred when available as they are exact and
continuous in space and time. The focus of this paper is analytical solutions for layered

* Elliot J. Carr
elliot.carr@qut.edu.au
1
School of Mathematical Sciences, Queensland University of Technology (QUT), Brisbane,
Australia

13
Vol.:(0123456789)
 E. J. Carr
40

porous media, which are commonly observed in natural and constructed environments
such as stratified soils and landfill clay liners (Liu et  al. 1998). For this problem, the
transport coefficients (dispersion coefficient, retardation factor, etc.) are piecewise con-
stant with the goal being to obtain the solute concentration in each layer as a function of
space and time (Fig. 1).
Analytical solutions of advection–dispersion–reaction equations (and closely related
equations such as advection–diffusion and reaction–diffusion equations) in layered media
continue to attract interest (Zimmerman et  al. 2016; Yang and Liu 2017; Guerrero et  al.
2013; Carr and Pontrelli 2018; Rodrigo and Worthy 2016). Mainly due to their prevalence
in heat conduction problems, most analytical solutions in the literature are developed for
multilayer diffusion problems (Carr and Turner 2016; Carr and Pontrelli 2018; Sun and
Wichman 2004; de  Monte 2002; Hickson et  al. 2009; Rodrigo and Worthy 2016) with
significantly less literature concerning advection–dispersion, reaction–diffusion or advec-
tion–dispersion–reaction problems.
Analytical solutions for advection–dispersion equations in layered media have been pre-
sented by a limited number of authors. Amongst the earliest work in this area is that of Leij
et al. (1991), who applied the Laplace transform to solve the advection–dispersion equation
(with retardation factor) on a semi-infinite two-layer medium with finite first layer, semi-
infinite second layer and continuity of concentration and dispersive flux at the interfaces
between adjacent layers. Both concentration-type and flux-type boundary conditions were
considered at the inlet, and a zero concentration gradient was applied at the outlet (Fig. 1).
Exact expressions for the concentration in the Laplace domain were obtained and numeri-
cally inverted.
In follow-up work, Leij and van Genuchten (1995) derived approximate analytical solu-
tions by first expanding the Laplace domain concentration in both layers as infinite series,
truncating each series after the first term and employing analytical inversion of the Laplace
transform to convert the concentration back to the time domain. Subsequent analytical
solutions for finite layered media and an arbitrary number of layers were derived by Liu

Fig. 1  Advection–dispersion–reaction in an m-layered medium. We solve for the concentration c(x,  t),

where the retardation factor R, dispersion coefficient D, pore water velocity v, volumetric water content 𝜃
and rate constants for first-order decay 𝜇 and zero-order production 𝛾 are constant within each layer but vary
across layers. Continuity of concentration and dispersive flux are imposed at the interfaces between adja-
cent layers ( x = 𝓁i , i = 1, … , m − 1 ), and transient boundary conditions of arbitrary type are specified at the
inlet ( x = 0 ) and outlet ( x = L)

13
New Semi-Analytical Solutions for Advection–Dispersion… 41

et  al. (1998) and then later by Guerrero et  al. (2013), both using the method of eigen-
function expansion. Liu et  al. (1998) considered the advection–dispersion equation and
Guerrero et  al. (2013) the advection–diffusion–reaction equation with a first-order decay
term. Both papers treated inlet boundary conditions of flux type only with Liu et al. (1998)
allowing for an arbitrary time-varying inlet concentration and Guerrero et al. (2013) a con-
stant inlet concentration. Both approaches require a transcendental equation to be solved
numerically for the eigenvalues with the accuracy of the solutions depending on the num-
ber of eigenvalues used in the expansions.
Recent work has revisited the use of Laplace transforms for solving multilayer transport
problems due to the ease of treating different boundary conditions and the ability to avoid
solving complicated nonlinear transcendental equations as required by eigenfunction expan-
sion solutions (Carr and Turner 2016). Namely, Carr and Turner (2016) and Rodrigo and
Worthy (2016) solved the multilayer diffusion problem using the Laplace transform for an
arbitrary number of layers and various types of boundary and interface conditions. Zimmer-
man et  al. (2016) solved a reaction–diffusion equation (without retardation factor) with a
first-order reaction term on a finite medium consisting of an arbitrary number of layers. All
three of these papers (Carr and Turner 2016; Rodrigo and Worthy 2016; Zimmerman et al.
2016) introduce an unknown function of time at each interface, representing either the dif-
fusive/dispersive flux or concentration, which allows the multilayer problem to be isolated
and solved separately on each layer in the Laplace domain before being inverted back to the
time domain.
Further relevant work involving the Laplace transform has focussed on the permeable reac-
tion barrier–aquifer problem. In particular, semi-analytical solutions for the advection–dis-
persion–reaction equation have been developed by Park and Zhan (2009) for a semi-infinite
one-dimensional medium using the Laplace transform and Chen et al. (2016) for a finite two-
dimensional medium using Laplace and finite Fourier transforms. Both solutions are limited
to two layers and numerically invert the Laplace transform using the de Hoog et al. (1982)
algorithm.
The aim of the current paper is to extend, generalize and merge the work of Carr
and Turner (2016), Rodrigo and Worthy (2016), Zimmerman et  al. (2016) and Park
and Zhan (2009) to solve advection–dispersion–reaction problems in one-dimensional
media comprising an arbitrary number of layers using the Laplace transform. The
derived solution is applicable to the most general form of linear advection–disper-
sion–reaction equation (van Genuchten and Alves 1982), a finite medium comprising an
arbitrary number of layers, continuity of concentration and dispersive flux at the inter-
faces between adjacent layers and transient boundary conditions of arbitrary type at the
inlet and outlet. To the best of our knowledge, analytical solutions satisfying all the
above conditions have not previously appeared in the published literature. The derived
solutions and resulting code supersede our previous work on multilayer diffusion (Carr
and Turner 2016; Carr and March 2018) by removing the requirement of numerically
computing eigenvalues in each layer.
The remaining sections of this paper are organized as follows: In Sect. 2, we describe
the multilayer advection–dispersion–reaction problem considered in this work. Section  3
develops our proposed solution procedure using the Laplace transform. In Sect.  4, the
developed solutions are applied to a wide selection of test cases and compared to other
analytical and numerical solutions available in the literature. In Sect. 5, we summarize the
work and discuss possible avenues for future research.

13
 E. J. Carr
42

2 Multilayer Transport Model

We consider solute transport across an m-layered porous medium partitioned as

0 = 𝓁0 < 𝓁1 < ⋯ < 𝓁m−1 < 𝓁m = L (Fig.  1). Let ci (x, t) be the solute concentration
[ ML−3 ] in the ith layer, where x ∈ (𝓁i−1 , 𝓁i ) is the distance from the inlet at x = 0 and
t > 0 is time. The governing transport equation in the ith layer is

𝜕ci 𝜕2 c 𝜕c
Ri = Di 2i − vi i − 𝜇i ci + 𝛾i , (1)
𝜕t 𝜕x 𝜕x
where Ri > 0 is the retardation factor [−] , Di > 0 is the dispersion coefficient [ L2 T−1 ], vi
is the pore water velocity [ LT−1 ], 𝜇i is the rate constant for first-order decay [ T−1 ] and 𝛾i is
the rate constant for zero-order production [ ML−3 T−1 ] (van Genuchten and Alves 1982).
Throughout this paper, we will denote by c(x, t) the result of amalgamating the layer con-
centrations, ci (x, t) ( i = 1, … , m ), as defined in Fig. 1.
The transport equations (1) are accompanied by initial, interface and boundary con-
ditions. The concentration is initially assumed constant in each layer
ci (x, 0) = fi , (2)
concentration and dispersive flux is assumed continuous at the interfaces between adjacent
layers ( i = 1, … , m − 1 ) (Leij et al. 1991; Liu et al. 1998; Guerrero et al. 2013):
ci (𝓁i , t) = ci+1 (𝓁i , t), (3)

𝜕ci 𝜕ci+1
𝜃i Di (𝓁 , t) = 𝜃i+1 Di+1 (𝓁i , t), (4)
𝜕x i 𝜕x
and general Robin boundary conditions are considered at the inlet ( x = 0 ) and outlet ( x = L):
𝜕c1
a0 c1 (0, t) − b0 (0, t) = g0 (t), (5)
𝜕x

𝜕cm
aL cm (L, t) + bL (L, t) = gL (t). (6)
𝜕x
The constant 𝜃i appearing in the interface condition (4) is the volumetric water content
[ L3 L−3 ] in the ith layer (van Genuchten and Alves 1982). In the boundary conditions
(5)–(6), a0 , b0 , aL and bL are constants and g0 (t) and gL (t) are arbitrary specified functions
of time with the subscripts 0 and L denoting the inlet ( x = 0 ) and outlet ( x = L ), respec-
tively (Fig. 1). We remark that b0 and bL are both nonnegative, at least one of a0 or b0 must
be nonzero and at least one of aL and bL must be nonzero.
For solute transport problems, commonly (Leij et al. 1991; Liu et al. 1998; Goltz and
Huang 2017) either a concentration-type boundary condition
c1 (0, t) = c0 (t), (7)
or a flux-type boundary condition

13
New Semi-Analytical Solutions for Advection–Dispersion… 43

𝜕c1
v1 c1 (0, t) − D1 (0, t) = v1 c0 (t), (8)
𝜕x
is applied at the inlet, where c0 (t) is the specified inlet concentration, while a zero concen-
tration gradient is applied at the outlet:
𝜕cm
(L, t) = 0. (9)
𝜕x
These boundary conditions are obtained from the general boundary conditions (5)–(6) by
setting a0 = 1 , b0 = 0 , g0 (t) = c0 (t) [Eq. (7)], a0 = v1 , b0 = D1 , g0 (t) = v1 c0 (t) [Eq. (8)] and
aL = 0 , bL = 1 , gL (t) = 0 [Eq. (9)], respectively.

3 Semi‑Analytical Solution

3.1 General Solution for m Layers in Laplace Space

To solve the multilayer transport model (1)–(6), we reformulate the model into m isolated
single-layer problems (Carr and Turner 2016; Rodrigo and Worthy 2016; Zimmerman et al.
2016). Introducing unknown functions of time, gi (t) ( i = 1, … , m − 1 ), to denote the follow-
ing scalar multiple of the (negative) dispersive flux at the layer interfaces (Carr and Turner
2016; Rodrigo and Worthy 2016):
𝜕ci
gi (t) = 𝜃i Di (𝓁 , t),
𝜕x i
yields the following equivalent form for the multilayer transport model (1)–(6):

First layer ( i = 1)

𝜕c1 𝜕2 c 𝜕c
R1 = D1 21 − v1 1 − 𝜇1 c1 + 𝛾1 , (10)
𝜕t 𝜕x 𝜕x

c1 (x, 0) = f1 , (11)

𝜕c1
a0 c1 (0, t) − b0 (0, t) = g0 (t), (12)
𝜕t

𝜕c1
𝜃1 D1 (𝓁 , t) = g1 (t), (13)
𝜕x 1
Middle layers ( i = 2, … , m − 1)

𝜕ci 𝜕2 c 𝜕c
Ri = Di 2i − vi i − 𝜇i ci + 𝛾i , (14)
𝜕t 𝜕x 𝜕x

ci (x, 0) = fi , (15)

13
 E. J. Carr
44

𝜕ci
𝜃i Di (𝓁 , t) = gi−1 (t), (16)
𝜕x i−1

𝜕ci
𝜃i Di (𝓁 , t) = gi (t), (17)
𝜕x i
Last layer ( i = m)

𝜕cm 𝜕2 c 𝜕c
Rm = Dm 2m − vm m − 𝜇m cm + 𝛾m , (18)
𝜕t 𝜕x 𝜕x

cm (x, 0) = fm , (19)

𝜕cm
𝜃m Dm (𝓁 , t) = gm−1 (t), (20)
𝜕x m−1

𝜕cm
aL cm (L, t) + bL (L, t) = gL (t), (21)
𝜕x
with each problem coupled together by imposing continuity of concentration at the inter-
faces between adjacent layers (3) (Carr and Turner 2016; Rodrigo and Worthy 2016; Carr
and March 2018).
We remark that the advection–dispersion–reaction equations (10), (14) and (18) can be
reduced to standard heat/diffusion equations via the change of variables:
( [ 2 ])
𝛾i vi x vi 𝜇i
ci (x, t) = + exp − + t ui (x, t). (22)
𝜇i 2Di 4Ri Di Ri

This seems to suggest that the transformed problem is amendable to solution by methods
designed for the multilayer diffusion problem presented in previous work (e.g. Carr and
Turner (2016), etc). However, this is not the case as the transformation (22) gives rise to
an advection term in the interface conditions (13), (16)–(17) and (20) for the transformed
variable ui (x, t) that would require the solutions to be redeveloped. Moreover, the transfor-
mation complicates the problem converting the initial conditions (11), (15), (19) from con-
stants to spatially dependent. We therefore do not invoke the transformation (22) and take
an alternative approach in this paper.
To solve the isolated single-layer problems (10)–(13), (14)–(17) and (18)–(21), we take
Laplace transforms yielding the boundary value problems:

First layer ( i = 1)

𝛾1
D1 C1�� − v1 C1� − (𝜇1 + R1 s)C1 = −R1 f1 − , (23)
s

a0 C1 (0, s) − b0 C1� (0, s) = G0 (s), (24)

𝜃1 D1 C1� (𝓁1 , s) = G1 (s), (25)

13
New Semi-Analytical Solutions for Advection–Dispersion… 45

Middle layers ( i = 2, … , m − 1)

𝛾i
Di Ci�� − vi Ci� − (𝜇i + Ri s)Ci = −Ri fi − , (26)
s

𝜃i Di Ci� (𝓁i−1 , s) = Gi−1 (s), (27)

𝜃i Di Ci� (𝓁i , s) = Gi (s), (28)

Last layer ( i = m)

𝛾m
Dm Cm�� − vm Cm� − (𝜇m + Rm s)Cm = −Rm fm − , (29)
s

𝜃m Dm Cm� (𝓁m−1 , s) = Gm−1 (s), (30)

aL Cm (L, s) + bL Cm� (L, s) = GL (s), (31)

where the prime notation ( ′  ) denotes a derivative with respect to x, Ci (x, s) = L{ci (x, t)}
denotes the Laplace transform of ci (x, t) with transformation variable s ∈ ℂ and
Gi (s) = L{gi (t)} for i = 1, … , m − 1 . Both G0 (s) = L{g0 (t)} and GL (s) = L{gL (t)} are
assumed to be able to be found analytically. The boundary value problems (23)–(25),
(26)–(28) and (29)–(31) all involve second-order constant coefficient differential equations,
which can be solved using standard techniques to give the following expressions for the
concentration in the Laplace domain:
C1 (x, s) = A1 (x, s)G0 (s) + B1 (x, s)G1 (s) + P1 (x, s), (32)

Ci (x, s) = Ai (x, s)Gi−1 (s) + Bi (x, s)Gi (s) + Pi (x, s), i = 2, … , m − 1, (33)

Cm (x, s) = Am (x, s)Gm−1 (s) + Bm (x, s)GL (s) + Pm (x, s), (34)
where the functions Pi , Ai and Bi ( i = 1, … , m ) are defined in Table 1.
With all other variables in the expressions (32)–(34) defined, to determine
G1 (s), … , Gm−1 (s)  , the Laplace transformations of the unknown interface functions
g1 (t), … , gm−1 (t) , we enforce continuity of concentration (3) at each interface in the Laplace
domain (Carr and Turner 2016; Rodrigo and Worthy 2016; Carr and March 2018):
Ci (𝓁i , s) = Ci+1 (𝓁i , s), i = 1, … , m − 1. (35)
Substituting
[ (32)–(34) ]into the system of equations (35) yields a linear system for
T
𝐱 = G1 (s), … , Gm−1 (s)  , expressible in matrix form as
𝐀𝐱 = 𝐛, (36)
where 𝐀 = (ai,j ) ∈ ℂ (m−1)×(m−1)
is a tridiagonal matrix and 𝐛 = (bi ) ∈ ℂ (m−1)
is a vector
with entries:

13
 E. J. Carr
46

Table 1  Definition of the functions Pi (x, s) , Ai (x, s) and Bi (x, s) ( i = 1, … , m ) appearing in the Laplace
domain concentrations (32)–(34)

First layer ( i = 1)

{ }
a0
P1 (x, s) = Ψ1 (s) + 𝛽 (s) 𝜆1,1 (s)Ψ1,2 (x, s) − 𝜆1,2 (s)Ψ1,2 (𝓁1 , s)Ψ1,1 (x, s) Ψ1 (s)
{ 1
}
A1 (x, s) = 𝛽 1(s) 𝜆1,2 (s)Ψ1,2 (𝓁1 , s)Ψ1,1 (x, s) − 𝜆1,1 (s)Ψ1,2 (x, s)
1
{[ ] [ ] }
B1 (x, s) = 𝜃 D 1𝛽 (s) a0 − b0 𝜆1,1 (s) Ψ1,1 (0, s)Ψ1,2 (x, s) − a0 − b0 𝜆1,2 (s) Ψ1,1 (x, s)
1 1 1

Where
𝛽1 (s) = [a0 − b0 𝜆1,1 (s)]𝜆1,2 (s) exp(−[𝜆1,1 (s) − 𝜆1,2 (s)]𝓁1 ) − [a0 − b0 𝜆1,2 (s)]𝜆1,1 (s)
Middle layers ( i = 2, … , m − 1)
Pi (x, s) = Ψi (s)
{ ]
Ai (x, s) = 𝜃 D 1𝛽 (s) 𝜆i,2 (s)Ψi,2 (𝓁i , s)Ψi,1 (x, s) − 𝜆i,1 (s)Ψi,2 (x, s)
i i i
{ }
Bi (x, s) = 𝜃 D 1𝛽 (s) 𝜆i,1 (s)Ψi,1 (𝓁i−1 , s)Ψi,2 (x, s) − 𝜆i,2 (s)Ψi,1 (x, s)
i i i

Where
{ }
𝛽i (s) = 𝜆i,1 (s)𝜆i,2 (s) exp(−[𝜆i,1 (s) − 𝜆i,2 (s)](𝓁i − 𝓁i−1 )) − 1
Last layer ( i = m)
{ }
a
Pm (x, s) = Ψm (s) + 𝛽 L(s) 𝜆m,2 (s)Ψm,1 (x, s) − 𝜆m,1 (s)Ψm,1 (𝓁m−1 , s)Ψm,2 (x, s) Ψm (s)
{m[ ] [ ] }
Am (x, s) = 𝜃 D 1𝛽 (s) aL + bL 𝜆m,2 (s) Ψm,2 (L, s)Ψm,1 (x, s) − aL + bL 𝜆m,1 (s) Ψm,2 (x, s)
{
m m m
}
Bm (x, s) = 𝛽 1(s) 𝜆m,1 (s)Ψm,1 (𝓁m−1 , s)Ψm,2 (x, s) − 𝜆m,2 (s)Ψm,1 (x, s)
m

Where
𝛽m (s) = [aL + bL 𝜆m,2 (s)]𝜆m,1 (s) exp(−[𝜆m,1 (s) − 𝜆m,2 (s)](𝓁m − 𝓁m−1 )) − [aL + bL 𝜆m,1 (s)]𝜆m,2 (s)
For all layers ( i = 1, … , m)
𝛾i s−1 +Ri fi
Ψi (s) = 𝜇i +Ri s
[ ]
Ψi,1 (x, s) = exp 𝜆i,1 (s)(x − 𝓁i ) recalling 𝓁m = L
[ ]
Ψi,2 (x, s) = exp 𝜆i,2 (s)(x − 𝓁i−1 ) recalling 𝓁0 = 0
√
vi + v2i +4Di (Ri s+𝜇i )
𝜆i,1 (s) = 2Di
√
vi − v2i +4Di (Ri s+𝜇i )
𝜆i,2 (s) = 2Di

a1,1 = B1 (𝓁1 , s) − A2 (𝓁1 , s),

a1,2 = −B2 (𝓁1 , s),
ai,i−1 = Ai (𝓁i , s), i = 2, … , m − 2,
ai,i = Bi (𝓁i , s) − Ai+1 (𝓁i , s), i = 2, … , m − 2,
ai,i+1 = −Bi+1 (𝓁i , s), i = 2, … , m − 2,
am−1,m−2 = Am−1 (𝓁m−1 , s),
am−1,m−1 = Bm−1 (𝓁m−1 , s) − Am (𝓁m−1 , s),
b1 = P2 (𝓁1 , s) − P1 (𝓁1 , s) − A1 (𝓁1 , s)G0 (s),
bi = Pi+1 (𝓁i , s) − Pi (𝓁i , s), i = 2, … , m − 2,
bm−1 = Pm (𝓁m−1 , s) − Pm−1 (𝓁m−1 , s) + Bm (𝓁m−1 , s)GL (s).

13
New Semi-Analytical Solutions for Advection–Dispersion… 47

Solving the linear system (36) allows the functions G1 (s), … , Gm−1 (s) to be computed and
hence the Laplace transform of the concentration (32)–(34) can be evaluated at any x and s
in the Laplace domain.

3.2 Simplification for Two Layers in Laplace Space

The formulation presented in the previous section breaks down for m = 2 layers due to
the assumption of middle layers. While the expressions for C1 (x, s) (32) and C2 (x, s) (34)
remain valid:
C1 (x, s) = A1 (x, s)G0 (s) + B1 (x, s)G1 (s) + P1 (x, s),
C2 (x, s) = A2 (x, s)G1 (s) + B2 (x, s)GL (s) + P2 (x, s),

the defined entries of the linear system (36) are no longer valid for m = 2 . In this case, the
linear system (36) reduces to a single equation that can be solved to yield:
[ ]
G1 (s) = (B1 (𝓁1 , s) − A2 (𝓁1 , s))−1 P2 (𝓁1 , s) − P1 (𝓁1 , s) − A1 (𝓁1 , s)G0 (s) + B2 (𝓁1 , s)GL (s) ,

which allows C1 (x, s) and C2 (x, s) to be evaluated at any x and s.

3.3 Numerical Inversion of Laplace Transform

To convert the Laplace domain expressions (32)–(34) back to the time domain and
thereby obtain the concentration ci (x, t) ( i = 1, … , m ), we need to compute the inverse
Laplace transform:
{ }
2𝜋i ∫Γ
1
ci (x, t) = L−1 Ci (x, s) = est Ci (x, s) ds,

where Γ is a Hankel contour that begins at −∞ − 0i , winds around the origin and termi-
nates at −∞ + 0i (Trefethen et al. 2006). Introducing the change of variable z = st , using a
rational approximation to ez and applying residue calculus (see Trefethen et al. (2006) for
full details) yield the numerical inversion formula:
{ }
{ } 2 ∑
ci (x, t) = L−1 Ci (x, s) ≈ − ℜ wk Ci (x, sk ) , (37)
t k∈𝕆 N

where N is even, 𝕆N is the set of positive odd integers less than N  , sk = zk ∕t and wk , zk ∈ ℂ
are the residues and poles of the best (N, N) rational approximation to ez on the negative
real line. Both wk and zk are constants, which are independent of x and t and computed
using a supplied MATLAB function (Trefethen et al. 2006, Fig 4.1).

3.4 Treatment of Step Function Boundary Conditions

Suppose the inlet concentration c0 (t) in either the concentration-type (7) or flux-type (8)
boundary condition is a Heaviside step function of duration t0 > 0:

13
 E. J. Carr
48

{
c0 , 0 < t < t0 ,
c0 (t) = c0 H(t0 − t) =
0, t > t0 ,

where c0 is a constant. In solute transport problems, such boundary conditions are com-
monly (van Genuchten and Alves 1982; Leij et  al. 1991; Goltz and Huang 2017) paired
with a zero concentration gradient at the outlet (9) and lead to G0 (s) = exp(−t0 s)∕s and
G0 (s) = v1 exp(−t0 s)∕s in Eq. (32) for the concentration-type and flux-type boundary con-
dition, respectively. Such exponential functions are well known to cause numerical prob-
lems in algorithms for inverting Laplace transforms (Kuhlman 2013). The approximation
(37) indeed suffers from this issue as evaluating exp(−t0 s) at s = sk = zk ∕t for poles zk with
negative real part leads to floating point overflow for small t. To overcome this problem,
we use superposition of solutions. Consider, for example, the multilayer transport model
(1)–(4) subject to the boundary conditions (7) or (8), and (9). The solution to this problem
can be expressed as
{
ci (x, t),
� 0 < t < t0 ,
ci (x, t) =
ci (x, t) − �
� ci (x, t − t0 ), t > t0 ,

where ̃ ci (x, t) is the solution of the multilayer transport model (1)–(6) with g0 (t) = c0
and ̂ci (x, t) is the solution of the multilayer transport model (1)–(6) with g0 (t) = c0 ,
fi = 0 and 𝛾i = 0 . Both ̃ ci (x, t) and ̂
ci (x, t) are obtained using the approach outlined in
subsections 3.1–3.3.

4 Results

We now demonstrate application of our semi-analytical Laplace transform solution and

verify that it produces the correct results using a selection of test cases. The transport
parameters, initial conditions and boundary conditions for each problem are provided in
Tables 2 and 3. A MATLAB code implementing our semi-analytical solution and produc-
ing the results in this section is available for download from githu​b.com/ellio​tcarr​/Carr2​
020a.

4.1 One‑ and Two‑Layer Test Cases

Cases 1–4 consider advection–dispersion–reaction in a homogeneous (single-layer)

medium of length 30 cm . The concentration is initially zero everywhere, and a zero con-
centration gradient is applied at the outlet (9). Four different boundary conditions are con-
sidered at the inlet:

• Case 1: flux-type boundary condition (8) with constant inlet concentration c0 (t) = c0;
• Case 2: flux-type boundary condition (8) with step inlet concentration c0 (t) = c0 H(t0 − t)
and pulse duration t0 = 0.5 days;
• Case 3: concentration-type boundary condition (7) with constant inlet concentration
c0 (t) = c0;
• Case 4: concentration-type boundary condition (7) with step inlet concentration
c0 (t) = c0 H(t0 − t) and pulse duration t0 = 0.5 days.

13
New Semi-Analytical Solutions for Advection–Dispersion… 49

Table 2  Transport parameters, Case i 𝓁i Ri Di vi 𝜇i 𝛾i 𝜃i fi

geometry and initial conditions
for various test cases, where i is
1–4 1 10 1 50 75 2 1 0.4 0
the layer index and 𝓁i locates the
end of layer i [ cm] 2 30 1 50 75 2 1 0.4 0
5 1 10 1 50 25 0 0 0.4 0
2 30 1 20 40 0 0 0.25 0
6 1 10 1 50 25 0 0 0.4 0
2 30 1 20 40 0 0 0.25 0
7 1 10 1 50 25 0 0 0.4 0
2 30 1 20 40 0 0 0.25 0
8 1 10 3 50 25 3 0 0.4 0
2 20 2 20 40 4 0 0.25 0
9–10, 11 1 10 4.25 7 10 0 0 0.4 0
2 12 14 18 8 0 0 0.5 0
3 20 4.25 7 10 0 0 0.4 0
4 22 14 18 8 0 0 0.5 0
5 30 4.25 7 10 0 0 0.4 0
12 1 10 4.25 7 10 0 0 0.4 0
2 12 14 18 8 0 0 0.5 0
3 14 4.25 7 10 0 0 0.4 0
4 18 4.25 7 10 0 0 0.4 c0
5 20 4.25 7 10 0 0 0.4 0
6 22 14 18 8 0 0 0.5 0
7 30 4.25 7 10 0 0 0.4 0
13 1 10 4.25 7 10 3 2 0.4 0
2 12 14 18 8 2 4 0.5 0
3 20 4.25 7 10 3 2 0.4 0
4 22 14 18 8 2 4 0.5 c0
5 30 4.25 7 10 3 2 0.4 0

In the ith layer: Ri is the retardation factor [−] , Di is the dispersion

coefficient [ cm2 day−1 ], vi is the pore water velocity [ cm day−1 ], 𝜇i is
the rate constant for first-order decay [ day−1 ], 𝛾i is the rate constant for
zero-order production [ kg cm−3 day−1 ], 𝜃i is the volumetric water con-
tent [−] and fi is the initial constant concentration [ kg cm−3 ]. Layer 4
for case 12 is the artificial layer created to accommodate the injected
contaminant initial condition (Sect. 4.2)

Table 3  Choice of inlet boundary Case a0 b0 g0 (t)

condition (5) for various test
cases
1, 5–9 v1 D1 v1 c0
2, 10, 13 v1 D1 v1 c0 H(t0 − t)
3 1 0 c0
4 1 0 c0 H(t0 − t)
11 1 0 c0 𝛼te−𝛽t
12 0 1 0

For all test cases, a zero concentration gradient is assumed at the out-
let with parameters aL = 0 , bL = 1 and gL (t) = 0 specified in the outlet
boundary condition (6)

13
 E. J. Carr
50

To solve these single-layer problems using our semi-analytical method, we choose m = 2

layers and set the transport parameters equal in both layers (Table 2). In Table 4, the rela-
tive concentration distributions ( c(x, t)∕c0 ) obtained are compared to corresponding dis-
tributions obtained using analytical solutions given in the literature (van Genuchten and
Alves 1982, sections C7 and C8), which are valid for homogeneous (single-layer) media
only. For all four problems, the maximum absolute difference between the relative con-
centration distributions obtained using both solution methods is tabulated in Table  4 at
t = 10−3 , 0.1, 0.6, 1, 2, 4 days . These results demonstrate that both solutions are in excellent
agreement for each choice of inlet condition and verify that our semi-analytical solution
produces the correct results for single-layer media.
Next, we present results for three two-layer test cases that have frequently appeared in
the literature (Leij and van Genuchten 1995; Liu et  al. 1998; Guerrero et  al. 2013). The
three test cases, labelled cases 5–7, consider advection–dispersion (without decay or pro-
duction) in a medium of length 30 cm with first and second layers of length 10 cm and
20 cm , respectively. A constant flux-type boundary condition is applied at the inlet (Eq.
(8) with c0 (t) = c0 ), and a zero concentration gradient is applied at the outlet (9). Initially,
the concentration is assumed to be zero in both layers. Different transport parameter com-
binations are applied for the three test cases (Table 2). In Table 5, for all three test cases,
we report the relative concentration values obtained from our semi-analytical solution at
several equidistant points in space and time. These results are compared to those previously
reported by Guerrero et al. (2013), Liu et al. (1998) and Leij and van Genuchten (1995)
with all numerical values in Table 5 displayed to three decimal places to be consistent with
the numerical values reported in those papers. For all three test cases, our results agree pre-
cisely with those of Guerrero et al. (2013) to the precision reported. When compared to Liu
et al. (1998) and Leij and van Genuchten (1995), minor differences of ±0.001 are evident
as highlighted in Table 5. These discrepancies are likely to be explained by Leij and van
Genuchten (1995) considering a semi-infinite second layer and using a different method
for numerically inverting the Laplace transform and Liu et al. (1998) considering a finite
second layer of undisclosed length.
Lastly, we present results for a final two-layer test case previously considered by
Guerrero et  al. (2013). This problem, labelled case 8, is the same as case 6 with the
exception that first-order decay is present in both layers (Table  2). In Fig.  2, we plot

Table 4  Comparison between our semi-analytical Laplace transform solution (Sect.  3) and the analytical
solutions catalogued in van Genuchten and Alves (1982) for the homogeneous medium test cases (1–4)
Case t = 10−3 t = 0.1 t = 0.6 t=1 t=2 t=4

1 4.11 × 10−14 5.53 × 10−10 5.84 × 10−08 9.38 × 10−09 4.75 × 10−09 6.10 × 10−10

2 4.11 × 10−14 5.53 × 10−10 1.89 × 10−08 7.10 × 10−08 1.74 × 10−08 5.90 × 10−10
3 1.98 × 10−14 7.61 × 10−10 1.93 × 10−08 1.53 × 10−08 1.20 × 10−09 3.34 × 10−10
4 1.98 × 10−14 7.61 × 10−10 1.93 × 10−08 1.57 × 10−09 1.15 × 10−08 8.10 × 10−10

The tabulated values are the maximum absolute difference between the values of the relative concentra-
tion ( 
c(x, t)∕c0 ) computed using the two approaches over the discrete range x = 0, 2, … , 20 cm , with
N = 14 poles/residues used for numerically inverting the Laplace transform (37). As recommended by van
Genuchten and Alves (1982); if v1 L∕D1 > min(5 + 40v1 t∕(R1 L), 5 + 40v1 (t − t0 )∕(R1 L), 100) , the analytical
solution is calculated using the approximate solutions catalogued (van Genuchten and Alves 1982). Other-
wise, the eigenfunction expansion solutions are used with 1000 eigenvalues/terms taken in the series expan-
sions (see van Genuchten and Alves (1982) and our code for full details)

13
 Table 5  Relative concentration values ( c(x, t)∕c0 ) for test cases 5–7 computed using our semi-analytical Laplace transform (SALT) solution (Sect. 3) with N = 14 poles/resi-
dues used in the numerical inversion of the Laplace transform (37)
Case x t = 0.2 t = 0.4 t = 0.6 t = 0.8
SALT GPS LBE LVG SALT GPS LBE LVG SALT GPS LBE LVG SALT GPS LBE LVG

5 0 0.884 0.884 0.884 0.884 0.963 0.963 0.963 0.963 0.987 0.987 0.987 0.987 0.995 0.995 0.995 0.995
2 0.742 0.742 0.742 0.742 0.915 0.915 0.915 0.915 0.969 0.969 0.969 0.969 0.988 0.988 0.988 0.988
4 0.561 0.561 0.561 0.561 0.841 0.841 0.841 0.841 0.940 0.940 0.940 0.940 0.977 0.977 0.977 0.977
6 0.375 0.375 0.374 0.375 0.746 0.746 0.746 0.746 0.901 0.901 0.901 0.901 0.962 0.962 0.962 0.962
8 0.222 0.222 0.222 0.222 0.645 0.645 0.645 0.645 0.858 0.858 0.858 0.858 0.945 0.945 0.945 0.945
10 0.142 0.142 0.142 0.142 0.579 0.579 0.579 0.579 0.829 0.829 0.829 0.829 0.933 0.933 0.933 0.933
12 0.063 0.063 0.063 0.063 0.480 0.480 0.480 0.480 0.781 0.781 0.781 0.781 0.914 0.914 0.914 0.914
14 0.021 0.021 0.021 0.021 0.372 0.372 0.372 0.372 0.722 0.722 0.722 0.722 0.889 0.889 0.889 0.889
16 0.005 0.005 0.005 0.005 0.264 0.264 0.265 0.264 0.651 0.651 0.651 0.651 0.858 0.858 0.858 0.858
18 0.001 0.001 0.001 0.001 0.168 0.168 0.169 0.168 0.567 0.567 0.567 0.567 0.819 0.819 0.819 0.819
New Semi-Analytical Solutions for Advection–Dispersion…

20 0.000 0.000 0.000 0.000 0.094 0.094 0.094 0.094 0.473 0.473 0.473 0.473 0.770 0.770 0.770 0.770
6 0 0.978 0.978 0.977 0.978 0.998 0.998 0.998 0.998 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000
2 0.868 0.868 0.867 0.868 0.984 0.984 0.984 0.984 0.998 0.998 0.998 0.998 1.000 1.000 1.000 1.000
4 0.634 0.634 0.633 0.634 0.942 0.942 0.942 0.942 0.991 0.991 0.991 0.991 0.999 0.999 0.999 0.999
6 0.345 0.345 0.345 0.345 0.849 0.849 0.849 0.849 0.972 0.972 0.972 0.972 0.995 0.995 0.995 0.995
8 0.131 0.131 0.131 0.131 0.693 0.693 0.693 0.693 0.930 0.930 0.929 0.930 0.986 0.986 0.986 0.986
10 0.033 0.033 0.033 0.033 0.496 0.496 0.496 0.496 0.853 0.853 0.853 0.853 0.966 0.966 0.966 0.966
12 0.011 0.011 0.011 0.011 0.370 0.370 0.370 0.370 0.784 0.784 0.783 0.784 0.944 0.944 0.944 0.944
14 0.003 0.003 0.003 0.003 0.257 0.257 0.257 0.257 0.699 0.699 0.698 0.699 0.913 0.913 0.913 0.913
16 0.001 0.001 0.001 0.001 0.166 0.166 0.166 0.166 0.601 0.601 0.601 0.601 0.871 0.871 0.871 0.871
18 0.000 0.000 0.000 0.000 0.098 0.098 0.099 0.098 0.498 0.498 0.498 0.498 0.817 0.817 0.817 0.817
20 0.000 0.000 0.000 0.000 0.054 0.054 0.054 0.054 0.395 0.395 0.395 0.395 0.751 0.751 0.750 0.751
51

13
 Table 5  (continued)
52

Case x t = 0.2 t = 0.4 t = 0.6 t = 0.8

SALT GPS LBE LVG SALT GPS LBE LVG SALT GPS LBE LVG SALT GPS LBE LVG

13
7 0 0.999 0.999 0.999 0.999 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000
2 0.988 0.988 0.987 0.988 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000
4 0.928 0.928 0.928 0.928 0.999 0.999 0.999 0.999 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000
6 0.764 0.764 0.763 0.764 0.995 0.995 0.995 0.995 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000
8 0.496 0.496 0.495 0.496 0.976 0.976 0.976 0.976 0.998 0.998 0.998 0.998 0.999 0.999 0.999 0.999
10 0.152 0.152 0.152 0.152 0.780 0.780 0.779 0.780 0.940 0.940 0.939 0.940 0.979 0.979 0.978 0.979
12 0.049 0.049 0.050 0.049 0.600 0.600 0.600 0.600 0.870 0.870 0.870 0.870 0.952 0.952 0.952 0.952
14 0.013 0.013 0.013 0.013 0.418 0.418 0.418 0.417 0.773 0.773 0.773 0.773 0.911 0.911 0.910 0.911
16 0.003 0.003 0.003 0.003 0.262 0.262 0.262 0.262 0.653 0.653 0.653 0.653 0.851 0.851 0.851 0.851
18 0.000 0.000 0.000 0.000 0.148 0.148 0.148 0.148 0.522 0.522 0.522 0.522 0.774 0.774 0.774 0.774
20 0.000 0.000 0.000 0.000 0.075 0.075 0.075 0.075 0.393 0.393 0.393 0.393 0.681 0.681 0.681 0.681

Results are compared with previously published values given by Guerrero et al. (2013) (GPS), Liu et al. (1998) (LBE) and Leij and van Genuchten (1995) (LVG). Bold cells
highlight discrepancies between the different methods by indicating when one of the four computed values differs from the others
E. J. Carr
New Semi-Analytical Solutions for Advection–Dispersion… 53

Fig. 2  Relative concentration spatial profiles ( c(x, t)∕c0 ) over time for test case 8. Continuous lines indicate
our semi-analytical Laplace transform solution (Sect.  3) with N = 14 poles/residues used in the numeri-
cal inversion of the Laplace transform (37). Circle markers indicate the numerical solution obtained using
the finite volume scheme outlined in “Appendix” computed with n = 601 nodes but plotted with markers
at x = 0, 1, … , 20 cm only. Crosses indicate the exact steady-state solution obtained by solving the steady-
state analogue of the multilayer transport model (1)–(6) analytically. The vertical dashed line indicates the
location of the interface at x = 10 cm . The plot is truncated at x = 20 cm rather than shown for the full
length of the medium ( x = 30 cm ) inline with Guerrero et al. (2013)

the relative concentration distributions obtained from our semi-analytical solution at

t = 0.2, 0.4, 0.6, 0.8 days (as in Guerrero et  al. (2013)) and t = 103 when the solution is
visible indistinguishable from its steady state. At short time, our results match well with
Guerrero et  al. (2013)’s results but differ for long times and at steady state. However,
agreement with the numerical solution (outlined in “Appendix”) as well as the exact
steady-state solution obtained by solving the steady-state analogue of the multilayer
transport model (1)–(6) (Fig. 2) supports that our approach produces the correct results.

4.2 Multiple Layer Test Cases

We now apply our semi-analytical solution to some test cases involving five or more
layers. Our solutions are compared to previously reported results from the literature
and/or numerical solutions obtained using a finite volume method, briefly discussed in
“Appendix”.
Firstly, we consider the five-layer test case previously considered by Liu et  al.
(1998) and Guerrero et  al. (2013). This test case involves advection–dispersion (with-
out decay or production) in a medium of length 30 cm consisting of alternating layers of
sand–clay–sand–clay–sand (cases 9–10 in Tables 2, 3). The initial concentration is zero in
each layer, and a zero concentration gradient is applied at the outlet. Two different bound-
ary conditions are considered at the inlet:

• Case 9: flux-type boundary condition (8) with constant inlet concentration c0 (t) = c0;
• Case 10: flux-type boundary condition (8) with step inlet concentration c0 (t) = c0 H(t0 − t)
and pulse duration t0 = 3 days.

13
 E. J. Carr
54

Fig. 3  Relative concentration spatial profiles ( c(x, t)∕c0 ) over time for test cases 9–13. Continuous lines
indicate our semi-analytical Laplace transform solution (Sect.  3) with N = 14 poles/residues used in the
numerical inversion of the Laplace transform (37). Circle markers indicate the numerical solution obtained
using the finite volume scheme outlined in “Appendix” computed with n = 601 nodes but plotted with
markers at x = 0, 1, … , 30 cm only. Vertical dashed lines indicate the location of the layer interfaces at
x = 10, 12, 20, 22 cm

In Fig.  3, we plot the relative concentration distributions ( c(x, t)∕c0 ) obtained from our
semi-analytical solution at several points in time. Only case 9 is considered by Liu et al.
(1998) and Guerrero et  al. (2013) who both report relative concentration distributions at
t = 2, 6, 10 days only with their solutions at these times all in excellent agreement with
those for case 9 in Fig. 3. For case 10 and the remaining times for case 9, agreement with
the relative concentration distributions computed using the numerical method discussed
in “Appendix” demonstrates that our semi-analytical solution produces the correct results.
To further highlight the capability of our approach, we consider three final test cases
(cases 11–13 in Tables 2, 3). Case 11 is case 9 with a continuously varying inlet concen-
tration (Table  3). Case 12 considers the same five-layer medium as cases 9–11 but with
a zero concentration gradient applied at both the inlet and outlet and an initial condition
modelling injection of a contaminant into the medium at time t = 0 between x = 14 cm and

13
New Semi-Analytical Solutions for Advection–Dispersion… 55

x = 18 cm (Goltz and Huang 2017). This test case is solved by creating an artificial layer
extending from x = 14 cm to x = 18 cm having the same transport parameters as the sand
layer in which it is located, leading to a seven-layer formulation as described in Table 2.
Finally, case 13 demonstrates the full capability of our semi-analytical solution with
nonzero rate constants of decay and production in each layer and multiple contaminant
sources (Tables 2, 3). The agreement with the relative concentration distributions obtained
using the numerical method evident in Fig. 3 for cases 11–13 further confirms the correct-
ness of our semi-analytical solution approach.

5 Conclusion

In this paper, we have developed a semi-analytical Laplace transform-based method to

solve the one-dimensional linear advection–dispersion–reaction equation in a layered
medium. The novelty of the approach is to introduce unknown functions at the interfaces
between adjacent layers, which allows the multilayer problem to be isolated and solved
separately on each layer before being numerically inverted back to the time domain. Our
derived solution is quite general in that it can be applied to problems involving an arbitrary
number of layers and arbitrary time-varying boundary conditions at the inlet and outlet.
The derived solutions extend and generalize recent work on diffusion (Carr and Turner
2016; Carr and March 2018; Rodrigo and Worthy 2016) and reaction–diffusion (Zimmer-
man et al. 2016) in layered media.
The solutions presented in this paper, and our MATLAB code, are limited to constant initial
conditions in each layer and interface conditions imposing continuity of concentration and dis-
persive flux between adjacent layers. However, extension to other interface conditions that do
not impose continuity of concentration (see e.g. Carr and March 2018), is straightforward and
achieved in our formulation by replacing Eq. (35) with the Laplace transform of the imposed
condition. For example, solving the multilayer transport model (1)–(6) with concentration
continuity (3) replaced by the partition interface condition (Carr and March 2018; Carr and
Pontrelli 2018), ci (𝓁i , t) = 𝛼i ci+1 (𝓁i , t) where 𝛼i > 0 is a specified constant, simply requires
replacement of Eq. (35) with Ci (𝓁i , s) = 𝛼i Ci+1 (𝓁i , s) and ultimately a small modification to
the entries of the linear system (36). Treatment of spatially varying initial conditions, where fi
is now fi (x) in the boundary value problems (23)–(25), (26)–(28) and (29)–(31), is more chal-
lenging but possible and would lead to more complicated expressions for the Laplace domain
concentration than those defined in Eqs. (32)–(34) and Table 1. The method for inverting the
Laplace transform can lead to unreliable results for advection-dominated transport so address-
ing this limitation would be valuable to pursue in the future. Another possible avenue for
future research is to extend the semi-analytical solutions to accommodate multispecies trans-
port and/or rate-limited sorption, as described by Chen et al. (2019a, 2019b) for single-layer
media and Mieles and Zhan (2012) and Chen et al. (2018) for two-layer media. We address the
case of multispecies multilayer transport in a forthcoming paper (Carr 2020).

Appendix: Numerical Solution

In this appendix, we briefly outline the finite volume scheme used to obtain a numerical solu-
tion to the multilayer transport model (1)–(6), as mentioned in Sect. 4.

13
 E. J. Carr
56

The interval [0, L] is discretized using a uniform grid consisting of n nodes with the kth
node located at x = (k − 1)h =∶ xk , where k = 1, … , n and h = L∕(n − 1) . The number of
nodes n is chosen to ensure that a node coincides with each interface ( x = 𝓁i , i = 1, … , m − 1 ).
Let ck (t) denote the numerical approximation to c(x, t) at x = xk and
ck − ck−1 c + ck
Ji,k = Di − vi k−1 ,
h 2
Si,k = −𝜇i ck + 𝛾i .

The discrete system takes the form:

d𝐜
𝐌 = 𝐅(𝐜), 𝐜(0) = 𝐜0 , (38)
dt
where 𝐌 ∈ ℝn×n , 𝐜 = (c1 , … , cn )T ∈ ℝn and 𝐅 = (F1 , … , Fn )T ∈ ℝn . The initial solution
vector 𝐜0 ∈ ℝn gets its entries from the initial condition (2) with first entry f1 , last entry fm
and kth entry ( k = 2, … , n − 1 ) equal to fi if xk ∈ (𝓁i−1 , 𝓁i ) or (fi + fi+1 )∕2 if xk = 𝓁i , where
i = 1, … , m − 1 is the interface index. The form of 𝐌 depends on the choice of boundary
conditions at the inlet and outlet:
≠ 0 and bL ≠ 0,
= 0 and bL ≠ 0,
⎧ 𝐈, if b0
⎪ 𝐈 − 𝐞1 𝐞T1 ,
≠ 0 and bL = 0,
if b0
𝐌=⎨
𝐈 − 𝐞n 𝐞Tn , if b0
⎪
⎩ 𝐈 − 𝐞1 𝐞T1 − 𝐞n 𝐞Tn , if b0 = 0 and bL = 0,

where 𝐈 is the n × n identity matrix and 𝐞k is the kth column of 𝐈 . The components of 𝐅 are
defined as follows:
F1 = a0 c1 − g0 (t)

if b0 = 0,
D1 a
J1,2 + g (t)
b0 0
+ (v1 − D1 b0 )c1 + 2h S1,1
0
F1 = h
R
2 1

if b0 ≠ 0,
Ji,k+1 − Ji,k + hSi,k
Fk =
hRi

if xk ∈ (𝓁i−1 , 𝓁i ),
( )
𝜃i+1 Ji+1,k+1 − 𝜃i Ji,k + h2 𝜃i Si,k + 𝜃i+1 Si+1,k
h( )
Fk =
2
𝜃i Ri + 𝜃i+1 Ri+1

if xk = 𝓁i and i = 2, … , m − 1,
Fn = aL cn − gL (t)

if bL = 0 , and

13
New Semi-Analytical Solutions for Advection–Dispersion… 57

Dm a
g (t)
bL L
− (vm + Dm bL )cn − Jm,n + h2 Sm,n
L
Fn = h
R
2 m

if bL ≠ 0 . The initial value problem (38) is solved using MATLAB’s in-built ode15s
solver with the default tolerances and options of 𝙼𝚊𝚜𝚜 = 𝐌 and 𝙼𝚊𝚜𝚜𝚂𝚒𝚗𝚐𝚞𝚕𝚊𝚛 = 𝚝𝚛𝚞𝚎 if
b0 and/or bL is zero. The interval of integration (tspan) is chosen to return the solution at
the appropriate times shown in Figs. 2 and 3.

References
Carr, E.J.: Generalized semi-analytical solution for coupled multispecies advection–dispersion equations in
multilayer porous media (2020). arXiv​:2006.15793​
Carr, E.J., March, N.G.: Semi-analytical solution of multilayer diffusion problems with time-varying bound-
ary conditions and general interface conditions. Appl. Math. Comput. 333, 286–303 (2018)
Carr, E.J., Pontrelli, G.: Modelling mass diffusion for a multi-layer sphere immersed in a semi-infinite
medium: application to drug delivery. Math. Biosci. 303, 1–9 (2018)
Carr, E.J., Turner, I.W.: A semi-analytical solution for multilayer diffusion in a composite medium consist-
ing of a large number of layers. Appl. Math. Model. 40, 7034–7050 (2016)
Chen, J.S., Hsu, S.Y., Li, M.H., Liu, C.W.: Assessing the performance of a permeable reactive barrier–aqui-
fer system using a dual-domain solute transport model. J. Hydrol. 543, 849–860 (2016)
Chen, H., Park, E., Hu, C.: A design solution of PRB with multispecies transport based on a multidomain
system. Environ. Earth Sci. 77, 630 (2018)
Chen, J.S., Ho, Y.C., Liang, C.P., Wang, S.W., Liu, C.W.: Semi-analytical model for coupled multispecies
advective–dispersive transport subject to rate-limited sorption. J. Hydrol. 579, 124164 (2019a)
Chen, J.S., Liang, C.P., Chang, C.H., Wan, M.H.: Simulating three-dimensional plume migration of a radio-
nuclide decay chain through groundwater. Energies 12, 3740 (2019b)
de Hoog, F.R., Knight, J.H., Stokes, A.N.: An improved method for numerical inversion of Laplace trans-
forms. SIAM J. Sci. Stat. Comput. 3, 357–366 (1982)
de Monte, F.: An analytic approach to the unsteady heat conduction process in one-dimensional composite
media. Int. J. Heat Mass Transf. 45, 1333–1343 (2002)
Goltz, M., Huang, J.: Analytical Modeling of Solute Transport in Groundwater. Wiley, Hoboken (2017)
Guerrero, J.S.P., Pimentel, L.C.G., Skaggs, T.H.: Analytical solution for the advection–dispersion transport
equation in layered media. Int. J. Heat Mass Transf. 56, 274–282 (2013)
Hickson, R.I., Barry, S.I., Mercer, G.N.: Critical times in multilayer diffusion. Part 1: exact solutions. Int. J.
Heat Mass Transf. 52, 5776–5783 (2009)
Kuhlman, K.L.: Review of inverse Laplace transform algorithms for Laplace-space numerical approaches.
Numer. Algorithms 63, 339–355 (2013)
Leij, F.J., van Genuchten, M.T.: Approximate analytical solutions for solute transport in two-layer porous
media. Transp. Porous Med. 18, 65–85 (1995)
Leij, F.J., Dane, J.H., van Genuchten, M.T.: Mathematical analysis of one-dimensional solute transport in a
layered soil profile. Soil Sci. Soc. Am. J. 55, 944–953 (1991)
Liu, C., Ball, W.P., Ellis, J.H.: An analytical solution to the one-dimensional solute advection–dispersion
equation in multi-layer porous media. Transp. Porous Med. 30, 25–43 (1998)
Mieles, J., Zhan, H.: Analytical solutions of one-dimensional multispecies reactive transport in a permeable
reactive barrier–aquifer system. J. Contam. Hydrol. 134–135, 54–68 (2012)
Park, E., Zhan, H.: One-dimensional solute transport in a permeable reactive barrier–aquifer system. Water
Resour. Res. 45, W07502 (2009)
Rodrigo, M.R., Worthy, A.L.: Solution of multilayer diffusion problems via the Laplace transform. J. Math.
Anal. Appl. 444, 475–502 (2016)
Sun, Y., Wichman, I.S.: On transient heat conduction in a one-dimensional composite slab. Int. J. Heat Mass
Transf. 47, 1555–1559 (2004)
Trefethen, L.N., Weideman, J.A.C., Schmelzer, T.: Talbot quadratures and rational approximations. BIT
Numer. Math. 46, 653–670 (2006)
van Genuchten, M.T., Alves, W.J.: Analytical solutions of the one-dimensional convective-dispersive solute
transport equation. US Department of Agriculture p Technical Bulletin No. 1661 (1982)

13
 E. J. Carr
58

Yang, B., Liu, S.: Closed-form analytical solutions of transient heat conduction in hollow composite cylin-
ders with any number of layers. Int. J. Heat Mass Transf. 108, 907–917 (2017)
Zimmerman, R.A., Jankowski, T.A., Tartakovsky, D.M.: Analytical models of axisymmetric reaction–diffu-
sion phenomena in composite media. Int. J. Heat Mass Transf. 99, 425–431 (2016)

Publisher’s Note  Springer Nature remains neutral with regard to jurisdictional claims in published maps and
institutional affiliations.

13